Torvalds is absolutely on point here, esp. at the end. It's the same in my circles as well - the brilliant ones (whom I've known for over 30 years) are also the ones who quickly picked up on LLM assistance.
@troed
The only issue I see with LLM use, is consistency and being deterministic.
If you give one prompt, will it always result in the same code?
If the prompt to the LLM becomes the new source code, it should be possible to have a reproducible built.
If AI can be coerced to do that, I think most aversions against the technology will vanish.
@gunstick Humans aren't deterministic and we've been able to work with those for a long time :) The source code the LLM outputs is still there if we want to rebuild the same thing.
@troed
How good is the source code?
Does it suffer from "reinvent the wheel" or does it extensively use existing libraries and tools?
You see, it has been a long time since I looked at AI generated code.
@gunstick If you use the same kind of software engineering principles as you do with humans the output is very similar. Good design phase, test driven development etc.
@troed
Oh that's where the "slop" comes in: People who have never coded generate some random apps.
I doubt that this happens on big projects.
@gunstick Exactly. I have my projects on our public git and I'd invite anyone to point out where the "slop" is ...
I've stopped looking at the code. I only deal in the functionality, tests, requirements etc now. The models - even the free local ones - have become incredibly capable the last few months.